1 /*
   2  * Copyright (c) 1997, 2016, Oracle and/or its affiliates. All rights reserved.
   3  * DO NOT ALTER OR REMOVE COPYRIGHT NOTICES OR THIS FILE HEADER.
   4  *
   5  * This code is free software; you can redistribute it and/or modify it
   6  * under the terms of the GNU General Public License version 2 only, as
   7  * published by the Free Software Foundation.
   8  *
   9  * This code is distributed in the hope that it will be useful, but WITHOUT
  10  * ANY WARRANTY; without even the implied warranty of MERCHANTABILITY or
  11  * FITNESS FOR A PARTICULAR PURPOSE.  See the GNU General Public License
  12  * version 2 for more details (a copy is included in the LICENSE file that
  13  * accompanied this code).
  14  *
  15  * You should have received a copy of the GNU General Public License version
  16  * 2 along with this work; if not, write to the Free Software Foundation,
  17  * Inc., 51 Franklin St, Fifth Floor, Boston, MA 02110-1301 USA.
  18  *
  19  * Please contact Oracle, 500 Oracle Parkway, Redwood Shores, CA 94065 USA
  20  * or visit www.oracle.com if you need additional information or have any
  21  * questions.
  22  *
  23  */
  24 
  25 #include "precompiled.hpp"
  26 #include "classfile/altHashing.hpp"
  27 #include "classfile/compactHashtable.inline.hpp"
  28 #include "classfile/javaClasses.hpp"
  29 #include "classfile/symbolTable.hpp"
  30 #include "classfile/systemDictionary.hpp"
  31 #include "gc/shared/collectedHeap.inline.hpp"
  32 #include "gc/shared/gcLocker.inline.hpp"
  33 #include "memory/allocation.inline.hpp"
  34 #include "memory/filemap.hpp"
  35 #include "memory/resourceArea.hpp"
  36 #include "oops/oop.inline.hpp"
  37 #include "runtime/atomic.hpp"
  38 #include "runtime/mutexLocker.hpp"
  39 #include "utilities/hashtable.inline.hpp"
  40 
  41 // --------------------------------------------------------------------------
  42 // the number of buckets a thread claims
  43 const int ClaimChunkSize = 32;
  44 
  45 SymbolTable* SymbolTable::_the_table = NULL;
  46 // Static arena for symbols that are not deallocated
  47 Arena* SymbolTable::_arena = NULL;
  48 bool SymbolTable::_needs_rehashing = false;
  49 bool SymbolTable::_lookup_shared_first = false;
  50 
  51 CompactHashtable<Symbol*, char> SymbolTable::_shared_table;
  52 
  53 Symbol* SymbolTable::allocate_symbol(const u1* name, int len, bool c_heap, TRAPS) {
  54   assert (len <= Symbol::max_length(), "should be checked by caller");
  55 
  56   Symbol* sym;
  57 
  58   if (DumpSharedSpaces) {
  59     // Allocate all symbols to CLD shared metaspace
  60     sym = new (len, ClassLoaderData::the_null_class_loader_data(), THREAD) Symbol(name, len, PERM_REFCOUNT);
  61   } else if (c_heap) {
  62     // refcount starts as 1
  63     sym = new (len, THREAD) Symbol(name, len, 1);
  64     assert(sym != NULL, "new should call vm_exit_out_of_memory if C_HEAP is exhausted");
  65   } else {
  66     // Allocate to global arena
  67     sym = new (len, arena(), THREAD) Symbol(name, len, PERM_REFCOUNT);
  68   }
  69   return sym;
  70 }
  71 
  72 void SymbolTable::initialize_symbols(int arena_alloc_size) {
  73   // Initialize the arena for global symbols, size passed in depends on CDS.
  74   if (arena_alloc_size == 0) {
  75     _arena = new (mtSymbol) Arena(mtSymbol);
  76   } else {
  77     _arena = new (mtSymbol) Arena(mtSymbol, arena_alloc_size);
  78   }
  79 }
  80 
  81 // Call function for all symbols in the symbol table.
  82 void SymbolTable::symbols_do(SymbolClosure *cl) {
  83   // all symbols from shared table
  84   _shared_table.symbols_do(cl);
  85 
  86   // all symbols from the dynamic table
  87   const int n = the_table()->table_size();
  88   for (int i = 0; i < n; i++) {
  89     for (HashtableEntry<Symbol*, mtSymbol>* p = the_table()->bucket(i);
  90          p != NULL;
  91          p = p->next()) {
  92       cl->do_symbol(p->literal_addr());
  93     }
  94   }
  95 }
  96 
  97 int SymbolTable::_symbols_removed = 0;
  98 int SymbolTable::_symbols_counted = 0;
  99 volatile int SymbolTable::_parallel_claimed_idx = 0;
 100 
 101 void SymbolTable::buckets_unlink(int start_idx, int end_idx, int* processed, int* removed) {
 102   for (int i = start_idx; i < end_idx; ++i) {
 103     HashtableEntry<Symbol*, mtSymbol>** p = the_table()->bucket_addr(i);
 104     HashtableEntry<Symbol*, mtSymbol>* entry = the_table()->bucket(i);
 105     while (entry != NULL) {
 106       // Shared entries are normally at the end of the bucket and if we run into
 107       // a shared entry, then there is nothing more to remove. However, if we
 108       // have rehashed the table, then the shared entries are no longer at the
 109       // end of the bucket.
 110       if (entry->is_shared() && !use_alternate_hashcode()) {
 111         break;
 112       }
 113       Symbol* s = entry->literal();
 114       (*processed)++;
 115       assert(s != NULL, "just checking");
 116       // If reference count is zero, remove.
 117       if (s->refcount() == 0) {
 118         assert(!entry->is_shared(), "shared entries should be kept live");
 119         delete s;
 120         (*removed)++;
 121         *p = entry->next();
 122         the_table()->free_entry(entry);
 123       } else {
 124         p = entry->next_addr();
 125       }
 126       // get next entry
 127       entry = (HashtableEntry<Symbol*, mtSymbol>*)HashtableEntry<Symbol*, mtSymbol>::make_ptr(*p);
 128     }
 129   }
 130 }
 131 
 132 // Remove unreferenced symbols from the symbol table
 133 // This is done late during GC.
 134 void SymbolTable::unlink(int* processed, int* removed) {
 135   size_t memory_total = 0;
 136   buckets_unlink(0, the_table()->table_size(), processed, removed);
 137   _symbols_removed += *removed;
 138   _symbols_counted += *processed;
 139 }
 140 
 141 void SymbolTable::possibly_parallel_unlink(int* processed, int* removed) {
 142   const int limit = the_table()->table_size();
 143 
 144   size_t memory_total = 0;
 145 
 146   for (;;) {
 147     // Grab next set of buckets to scan
 148     int start_idx = Atomic::add(ClaimChunkSize, &_parallel_claimed_idx) - ClaimChunkSize;
 149     if (start_idx >= limit) {
 150       // End of table
 151       break;
 152     }
 153 
 154     int end_idx = MIN2(limit, start_idx + ClaimChunkSize);
 155     buckets_unlink(start_idx, end_idx, processed, removed);
 156   }
 157   Atomic::add(*processed, &_symbols_counted);
 158   Atomic::add(*removed, &_symbols_removed);
 159 }
 160 
 161 // Create a new table and using alternate hash code, populate the new table
 162 // with the existing strings.   Set flag to use the alternate hash code afterwards.
 163 void SymbolTable::rehash_table() {
 164   if (DumpSharedSpaces) {
 165     tty->print_cr("Warning: rehash_table should not be called while dumping archive");
 166     return;
 167   }
 168 
 169   assert(SafepointSynchronize::is_at_safepoint(), "must be at safepoint");
 170   // This should never happen with -Xshare:dump but it might in testing mode.
 171   if (DumpSharedSpaces) return;
 172   // Create a new symbol table
 173   SymbolTable* new_table = new SymbolTable();
 174 
 175   the_table()->move_to(new_table);
 176 
 177   // Delete the table and buckets (entries are reused in new table).
 178   delete _the_table;
 179   // Don't check if we need rehashing until the table gets unbalanced again.
 180   // Then rehash with a new global seed.
 181   _needs_rehashing = false;
 182   _the_table = new_table;
 183 }
 184 
 185 // Lookup a symbol in a bucket.
 186 
 187 Symbol* SymbolTable::lookup_dynamic(int index, const char* name,
 188                                     int len, unsigned int hash) {
 189   int count = 0;
 190   for (HashtableEntry<Symbol*, mtSymbol>* e = bucket(index); e != NULL; e = e->next()) {
 191     count++;  // count all entries in this bucket, not just ones with same hash
 192     if (e->hash() == hash) {
 193       Symbol* sym = e->literal();
 194       if (sym->equals(name, len)) {
 195         // something is referencing this symbol now.
 196         sym->increment_refcount();
 197         return sym;
 198       }
 199     }
 200   }
 201   // If the bucket size is too deep check if this hash code is insufficient.
 202   if (count >= rehash_count && !needs_rehashing()) {
 203     _needs_rehashing = check_rehash_table(count);
 204   }
 205   return NULL;
 206 }
 207 
 208 Symbol* SymbolTable::lookup_shared(const char* name,
 209                                    int len, unsigned int hash) {
 210   if (use_alternate_hashcode()) {
 211     // hash_code parameter may use alternate hashing algorithm but the shared table
 212     // always uses the same original hash code.
 213     hash = hash_shared_symbol(name, len);
 214   }
 215   return _shared_table.lookup(name, hash, len);
 216 }
 217 
 218 Symbol* SymbolTable::lookup(int index, const char* name,
 219                             int len, unsigned int hash) {
 220   Symbol* sym;
 221   if (_lookup_shared_first) {
 222     sym = lookup_shared(name, len, hash);
 223     if (sym != NULL) {
 224       return sym;
 225     }
 226     _lookup_shared_first = false;
 227     return lookup_dynamic(index, name, len, hash);
 228   } else {
 229     sym = lookup_dynamic(index, name, len, hash);
 230     if (sym != NULL) {
 231       return sym;
 232     }
 233     sym = lookup_shared(name, len, hash);
 234     if (sym != NULL) {
 235       _lookup_shared_first = true;
 236     }
 237     return sym;
 238   }
 239 }
 240 
 241 u4 SymbolTable::encode_shared(Symbol* sym) {
 242   assert(DumpSharedSpaces, "called only during dump time");
 243   uintx base_address = uintx(MetaspaceShared::shared_rs()->base());
 244   uintx offset = uintx(sym) - base_address;
 245   assert(offset < 0x7fffffff, "sanity");
 246   return u4(offset);
 247 }
 248 
 249 Symbol* SymbolTable::decode_shared(u4 offset) {
 250   assert(!DumpSharedSpaces, "called only during runtime");
 251   uintx base_address = _shared_table.base_address();
 252   Symbol* sym = (Symbol*)(base_address + offset);
 253 
 254 #ifndef PRODUCT
 255   const char* s = (const char*)sym->bytes();
 256   int len = sym->utf8_length();
 257   unsigned int hash = hash_symbol(s, len);
 258   assert(sym == lookup_shared(s, len, hash), "must be shared symbol");
 259 #endif
 260 
 261   return sym;
 262 }
 263 
 264 // Pick hashing algorithm.
 265 unsigned int SymbolTable::hash_symbol(const char* s, int len) {
 266   return use_alternate_hashcode() ?
 267            AltHashing::murmur3_32(seed(), (const jbyte*)s, len) :
 268            java_lang_String::hash_code((const jbyte*)s, len);
 269 }
 270 
 271 unsigned int SymbolTable::hash_shared_symbol(const char* s, int len) {
 272   return java_lang_String::hash_code((const jbyte*)s, len);
 273 }
 274 
 275 
 276 // We take care not to be blocking while holding the
 277 // SymbolTable_lock. Otherwise, the system might deadlock, since the
 278 // symboltable is used during compilation (VM_thread) The lock free
 279 // synchronization is simplified by the fact that we do not delete
 280 // entries in the symbol table during normal execution (only during
 281 // safepoints).
 282 
 283 Symbol* SymbolTable::lookup(const char* name, int len, TRAPS) {
 284   unsigned int hashValue = hash_symbol(name, len);
 285   int index = the_table()->hash_to_index(hashValue);
 286 
 287   Symbol* s = the_table()->lookup(index, name, len, hashValue);
 288 
 289   // Found
 290   if (s != NULL) return s;
 291 
 292   // Grab SymbolTable_lock first.
 293   MutexLocker ml(SymbolTable_lock, THREAD);
 294 
 295   // Otherwise, add to symbol to table
 296   return the_table()->basic_add(index, (u1*)name, len, hashValue, true, THREAD);
 297 }
 298 
 299 Symbol* SymbolTable::lookup(const Symbol* sym, int begin, int end, TRAPS) {
 300   char* buffer;
 301   int index, len;
 302   unsigned int hashValue;
 303   char* name;
 304   {
 305     debug_only(NoSafepointVerifier nsv;)
 306 
 307     name = (char*)sym->base() + begin;
 308     len = end - begin;
 309     hashValue = hash_symbol(name, len);
 310     index = the_table()->hash_to_index(hashValue);
 311     Symbol* s = the_table()->lookup(index, name, len, hashValue);
 312 
 313     // Found
 314     if (s != NULL) return s;
 315   }
 316 
 317   // Otherwise, add to symbol to table. Copy to a C string first.
 318   char stack_buf[128];
 319   ResourceMark rm(THREAD);
 320   if (len <= 128) {
 321     buffer = stack_buf;
 322   } else {
 323     buffer = NEW_RESOURCE_ARRAY_IN_THREAD(THREAD, char, len);
 324   }
 325   for (int i=0; i<len; i++) {
 326     buffer[i] = name[i];
 327   }
 328   // Make sure there is no safepoint in the code above since name can't move.
 329   // We can't include the code in NoSafepointVerifier because of the
 330   // ResourceMark.
 331 
 332   // Grab SymbolTable_lock first.
 333   MutexLocker ml(SymbolTable_lock, THREAD);
 334 
 335   return the_table()->basic_add(index, (u1*)buffer, len, hashValue, true, THREAD);
 336 }
 337 
 338 Symbol* SymbolTable::lookup_only(const char* name, int len,
 339                                    unsigned int& hash) {
 340   hash = hash_symbol(name, len);
 341   int index = the_table()->hash_to_index(hash);
 342 
 343   Symbol* s = the_table()->lookup(index, name, len, hash);
 344   return s;
 345 }
 346 
 347 // Look up the address of the literal in the SymbolTable for this Symbol*
 348 // Do not create any new symbols
 349 // Do not increment the reference count to keep this alive
 350 Symbol** SymbolTable::lookup_symbol_addr(Symbol* sym){
 351   unsigned int hash = hash_symbol((char*)sym->bytes(), sym->utf8_length());
 352   int index = the_table()->hash_to_index(hash);
 353 
 354   for (HashtableEntry<Symbol*, mtSymbol>* e = the_table()->bucket(index); e != NULL; e = e->next()) {
 355     if (e->hash() == hash) {
 356       Symbol* literal_sym = e->literal();
 357       if (sym == literal_sym) {
 358         return e->literal_addr();
 359       }
 360     }
 361   }
 362   return NULL;
 363 }
 364 
 365 // Suggestion: Push unicode-based lookup all the way into the hashing
 366 // and probing logic, so there is no need for convert_to_utf8 until
 367 // an actual new Symbol* is created.
 368 Symbol* SymbolTable::lookup_unicode(const jchar* name, int utf16_length, TRAPS) {
 369   int utf8_length = UNICODE::utf8_length((jchar*) name, utf16_length);
 370   char stack_buf[128];
 371   if (utf8_length < (int) sizeof(stack_buf)) {
 372     char* chars = stack_buf;
 373     UNICODE::convert_to_utf8(name, utf16_length, chars);
 374     return lookup(chars, utf8_length, THREAD);
 375   } else {
 376     ResourceMark rm(THREAD);
 377     char* chars = NEW_RESOURCE_ARRAY(char, utf8_length + 1);;
 378     UNICODE::convert_to_utf8(name, utf16_length, chars);
 379     return lookup(chars, utf8_length, THREAD);
 380   }
 381 }
 382 
 383 Symbol* SymbolTable::lookup_only_unicode(const jchar* name, int utf16_length,
 384                                            unsigned int& hash) {
 385   int utf8_length = UNICODE::utf8_length((jchar*) name, utf16_length);
 386   char stack_buf[128];
 387   if (utf8_length < (int) sizeof(stack_buf)) {
 388     char* chars = stack_buf;
 389     UNICODE::convert_to_utf8(name, utf16_length, chars);
 390     return lookup_only(chars, utf8_length, hash);
 391   } else {
 392     ResourceMark rm;
 393     char* chars = NEW_RESOURCE_ARRAY(char, utf8_length + 1);;
 394     UNICODE::convert_to_utf8(name, utf16_length, chars);
 395     return lookup_only(chars, utf8_length, hash);
 396   }
 397 }
 398 
 399 void SymbolTable::add(ClassLoaderData* loader_data, const constantPoolHandle& cp,
 400                       int names_count,
 401                       const char** names, int* lengths, int* cp_indices,
 402                       unsigned int* hashValues, TRAPS) {
 403   // Grab SymbolTable_lock first.
 404   MutexLocker ml(SymbolTable_lock, THREAD);
 405 
 406   SymbolTable* table = the_table();
 407   bool added = table->basic_add(loader_data, cp, names_count, names, lengths,
 408                                 cp_indices, hashValues, CHECK);
 409   if (!added) {
 410     // do it the hard way
 411     for (int i=0; i<names_count; i++) {
 412       int index = table->hash_to_index(hashValues[i]);
 413       bool c_heap = !loader_data->is_the_null_class_loader_data();
 414       Symbol* sym = table->basic_add(index, (u1*)names[i], lengths[i], hashValues[i], c_heap, CHECK);
 415       cp->symbol_at_put(cp_indices[i], sym);
 416     }
 417   }
 418 }
 419 
 420 Symbol* SymbolTable::new_permanent_symbol(const char* name, TRAPS) {
 421   unsigned int hash;
 422   Symbol* result = SymbolTable::lookup_only((char*)name, (int)strlen(name), hash);
 423   if (result != NULL) {
 424     return result;
 425   }
 426   // Grab SymbolTable_lock first.
 427   MutexLocker ml(SymbolTable_lock, THREAD);
 428 
 429   SymbolTable* table = the_table();
 430   int index = table->hash_to_index(hash);
 431   return table->basic_add(index, (u1*)name, (int)strlen(name), hash, false, THREAD);
 432 }
 433 
 434 Symbol* SymbolTable::basic_add(int index_arg, u1 *name, int len,
 435                                unsigned int hashValue_arg, bool c_heap, TRAPS) {
 436   assert(!Universe::heap()->is_in_reserved(name),
 437          "proposed name of symbol must be stable");
 438 
 439   // Don't allow symbols to be created which cannot fit in a Symbol*.
 440   if (len > Symbol::max_length()) {
 441     THROW_MSG_0(vmSymbols::java_lang_InternalError(),
 442                 "name is too long to represent");
 443   }
 444 
 445   // Cannot hit a safepoint in this function because the "this" pointer can move.
 446   NoSafepointVerifier nsv;
 447 
 448   // Check if the symbol table has been rehashed, if so, need to recalculate
 449   // the hash value and index.
 450   unsigned int hashValue;
 451   int index;
 452   if (use_alternate_hashcode()) {
 453     hashValue = hash_symbol((const char*)name, len);
 454     index = hash_to_index(hashValue);
 455   } else {
 456     hashValue = hashValue_arg;
 457     index = index_arg;
 458   }
 459 
 460   // Since look-up was done lock-free, we need to check if another
 461   // thread beat us in the race to insert the symbol.
 462   Symbol* test = lookup(index, (char*)name, len, hashValue);
 463   if (test != NULL) {
 464     // A race occurred and another thread introduced the symbol.
 465     assert(test->refcount() != 0, "lookup should have incremented the count");
 466     return test;
 467   }
 468 
 469   // Create a new symbol.
 470   Symbol* sym = allocate_symbol(name, len, c_heap, CHECK_NULL);
 471   assert(sym->equals((char*)name, len), "symbol must be properly initialized");
 472 
 473   HashtableEntry<Symbol*, mtSymbol>* entry = new_entry(hashValue, sym);
 474   add_entry(index, entry);
 475   return sym;
 476 }
 477 
 478 // This version of basic_add adds symbols in batch from the constant pool
 479 // parsing.
 480 bool SymbolTable::basic_add(ClassLoaderData* loader_data, const constantPoolHandle& cp,
 481                             int names_count,
 482                             const char** names, int* lengths,
 483                             int* cp_indices, unsigned int* hashValues,
 484                             TRAPS) {
 485 
 486   // Check symbol names are not too long.  If any are too long, don't add any.
 487   for (int i = 0; i< names_count; i++) {
 488     if (lengths[i] > Symbol::max_length()) {
 489       THROW_MSG_0(vmSymbols::java_lang_InternalError(),
 490                   "name is too long to represent");
 491     }
 492   }
 493 
 494   // Cannot hit a safepoint in this function because the "this" pointer can move.
 495   NoSafepointVerifier nsv;
 496 
 497   for (int i=0; i<names_count; i++) {
 498     // Check if the symbol table has been rehashed, if so, need to recalculate
 499     // the hash value.
 500     unsigned int hashValue;
 501     if (use_alternate_hashcode()) {
 502       hashValue = hash_symbol(names[i], lengths[i]);
 503     } else {
 504       hashValue = hashValues[i];
 505     }
 506     // Since look-up was done lock-free, we need to check if another
 507     // thread beat us in the race to insert the symbol.
 508     int index = hash_to_index(hashValue);
 509     Symbol* test = lookup(index, names[i], lengths[i], hashValue);
 510     if (test != NULL) {
 511       // A race occurred and another thread introduced the symbol, this one
 512       // will be dropped and collected. Use test instead.
 513       cp->symbol_at_put(cp_indices[i], test);
 514       assert(test->refcount() != 0, "lookup should have incremented the count");
 515     } else {
 516       // Create a new symbol.  The null class loader is never unloaded so these
 517       // are allocated specially in a permanent arena.
 518       bool c_heap = !loader_data->is_the_null_class_loader_data();
 519       Symbol* sym = allocate_symbol((const u1*)names[i], lengths[i], c_heap, CHECK_(false));
 520       assert(sym->equals(names[i], lengths[i]), "symbol must be properly initialized");  // why wouldn't it be???
 521       HashtableEntry<Symbol*, mtSymbol>* entry = new_entry(hashValue, sym);
 522       add_entry(index, entry);
 523       cp->symbol_at_put(cp_indices[i], sym);
 524     }
 525   }
 526   return true;
 527 }
 528 
 529 
 530 void SymbolTable::verify() {
 531   for (int i = 0; i < the_table()->table_size(); ++i) {
 532     HashtableEntry<Symbol*, mtSymbol>* p = the_table()->bucket(i);
 533     for ( ; p != NULL; p = p->next()) {
 534       Symbol* s = (Symbol*)(p->literal());
 535       guarantee(s != NULL, "symbol is NULL");
 536       unsigned int h = hash_symbol((char*)s->bytes(), s->utf8_length());
 537       guarantee(p->hash() == h, "broken hash in symbol table entry");
 538       guarantee(the_table()->hash_to_index(h) == i,
 539                 "wrong index in symbol table");
 540     }
 541   }
 542 }
 543 
 544 void SymbolTable::dump(outputStream* st, bool verbose) {
 545   if (!verbose) {
 546     the_table()->dump_table(st, "SymbolTable");
 547   } else {
 548     st->print_cr("VERSION: 1.0");
 549     for (int i = 0; i < the_table()->table_size(); ++i) {
 550       HashtableEntry<Symbol*, mtSymbol>* p = the_table()->bucket(i);
 551       for ( ; p != NULL; p = p->next()) {
 552         Symbol* s = (Symbol*)(p->literal());
 553         const char* utf8_string = (const char*)s->bytes();
 554         int utf8_length = s->utf8_length();
 555         st->print("%d %d: ", utf8_length, s->refcount());
 556         HashtableTextDump::put_utf8(st, utf8_string, utf8_length);
 557         st->cr();
 558       }
 559     }
 560   }
 561 }
 562 
 563 void SymbolTable::serialize(SerializeClosure* soc) {
 564 #if INCLUDE_CDS
 565   _shared_table.reset();
 566   if (soc->writing()) {
 567     int num_buckets = the_table()->number_of_entries() /
 568                             SharedSymbolTableBucketSize;
 569     CompactSymbolTableWriter writer(num_buckets,
 570                                     &MetaspaceShared::stats()->symbol);
 571     for (int i = 0; i < the_table()->table_size(); ++i) {
 572       HashtableEntry<Symbol*, mtSymbol>* p = the_table()->bucket(i);
 573       for ( ; p != NULL; p = p->next()) {
 574         Symbol* s = (Symbol*)(p->literal());
 575       unsigned int fixed_hash =  hash_shared_symbol((char*)s->bytes(), s->utf8_length());
 576         assert(fixed_hash == p->hash(), "must not rehash during dumping");
 577         writer.add(fixed_hash, s);
 578       }
 579     }
 580 
 581     writer.dump(&_shared_table);
 582   }
 583 
 584   _shared_table.set_type(CompactHashtable<Symbol*, char>::_symbol_table);
 585   _shared_table.serialize(soc);
 586 
 587   if (soc->writing()) {
 588     // Verify table is correct
 589     Symbol* sym = vmSymbols::java_lang_Object();
 590     const char* name = (const char*)sym->bytes();
 591     int len = sym->utf8_length();
 592     unsigned int hash = hash_symbol(name, len);
 593     assert(sym == _shared_table.lookup(name, hash, len), "sanity");
 594 
 595     // Sanity. Make sure we don't use the shared table at dump time
 596     _shared_table.reset();
 597   }
 598 #endif
 599 }
 600 
 601 //---------------------------------------------------------------------------
 602 // Non-product code
 603 
 604 #ifndef PRODUCT
 605 
 606 void SymbolTable::print_histogram() {
 607   MutexLocker ml(SymbolTable_lock);
 608   const int results_length = 100;
 609   int counts[results_length];
 610   int sizes[results_length];
 611   int i,j;
 612 
 613   // initialize results to zero
 614   for (j = 0; j < results_length; j++) {
 615     counts[j] = 0;
 616     sizes[j] = 0;
 617   }
 618 
 619   int total_size = 0;
 620   int total_count = 0;
 621   int total_length = 0;
 622   int max_length = 0;
 623   int out_of_range_count = 0;
 624   int out_of_range_size = 0;
 625   for (i = 0; i < the_table()->table_size(); i++) {
 626     HashtableEntry<Symbol*, mtSymbol>* p = the_table()->bucket(i);
 627     for ( ; p != NULL; p = p->next()) {
 628       int size = p->literal()->size();
 629       int len = p->literal()->utf8_length();
 630       if (len < results_length) {
 631         counts[len]++;
 632         sizes[len] += size;
 633       } else {
 634         out_of_range_count++;
 635         out_of_range_size += size;
 636       }
 637       total_count++;
 638       total_size += size;
 639       total_length += len;
 640       max_length = MAX2(max_length, len);
 641     }
 642   }
 643   tty->print_cr("Symbol Table Histogram:");
 644   tty->print_cr("  Total number of symbols  %7d", total_count);
 645   tty->print_cr("  Total size in memory     %7dK",
 646           (total_size*wordSize)/1024);
 647   tty->print_cr("  Total counted            %7d", _symbols_counted);
 648   tty->print_cr("  Total removed            %7d", _symbols_removed);
 649   if (_symbols_counted > 0) {
 650     tty->print_cr("  Percent removed          %3.2f",
 651           ((float)_symbols_removed/(float)_symbols_counted)* 100);
 652   }
 653   tty->print_cr("  Reference counts         %7d", Symbol::_total_count);
 654   tty->print_cr("  Symbol arena used        " SIZE_FORMAT_W(7) "K", arena()->used()/1024);
 655   tty->print_cr("  Symbol arena size        " SIZE_FORMAT_W(7) "K", arena()->size_in_bytes()/1024);
 656   tty->print_cr("  Total symbol length      %7d", total_length);
 657   tty->print_cr("  Maximum symbol length    %7d", max_length);
 658   tty->print_cr("  Average symbol length    %7.2f", ((float) total_length / (float) total_count));
 659   tty->print_cr("  Symbol length histogram:");
 660   tty->print_cr("    %6s %10s %10s", "Length", "#Symbols", "Size");
 661   for (i = 0; i < results_length; i++) {
 662     if (counts[i] > 0) {
 663       tty->print_cr("    %6d %10d %10dK", i, counts[i], (sizes[i]*wordSize)/1024);
 664     }
 665   }
 666   tty->print_cr("  >=%6d %10d %10dK\n", results_length,
 667           out_of_range_count, (out_of_range_size*wordSize)/1024);
 668 }
 669 
 670 void SymbolTable::print() {
 671   for (int i = 0; i < the_table()->table_size(); ++i) {
 672     HashtableEntry<Symbol*, mtSymbol>** p = the_table()->bucket_addr(i);
 673     HashtableEntry<Symbol*, mtSymbol>* entry = the_table()->bucket(i);
 674     if (entry != NULL) {
 675       while (entry != NULL) {
 676         tty->print(PTR_FORMAT " ", p2i(entry->literal()));
 677         entry->literal()->print();
 678         tty->print(" %d", entry->literal()->refcount());
 679         p = entry->next_addr();
 680         entry = (HashtableEntry<Symbol*, mtSymbol>*)HashtableEntry<Symbol*, mtSymbol>::make_ptr(*p);
 681       }
 682       tty->cr();
 683     }
 684   }
 685 }
 686 #endif // PRODUCT
 687 
 688 
 689 // Utility for dumping symbols
 690 SymboltableDCmd::SymboltableDCmd(outputStream* output, bool heap) :
 691                                  DCmdWithParser(output, heap),
 692   _verbose("-verbose", "Dump the content of each symbol in the table",
 693            "BOOLEAN", false, "false") {
 694   _dcmdparser.add_dcmd_option(&_verbose);
 695 }
 696 
 697 void SymboltableDCmd::execute(DCmdSource source, TRAPS) {
 698   VM_DumpHashtable dumper(output(), VM_DumpHashtable::DumpSymbols,
 699                          _verbose.value());
 700   VMThread::execute(&dumper);
 701 }
 702 
 703 int SymboltableDCmd::num_arguments() {
 704   ResourceMark rm;
 705   SymboltableDCmd* dcmd = new SymboltableDCmd(NULL, false);
 706   if (dcmd != NULL) {
 707     DCmdMark mark(dcmd);
 708     return dcmd->_dcmdparser.num_arguments();
 709   } else {
 710     return 0;
 711   }
 712 }